Long-horizon embodied planning is challenging because the world does not only change through an agent's actions: exogenous processes (e.g., water heating, dominoes cascading) unfold concurrently with the agent's actions. We propose a framework for abstract world models that jointly learns (i) symbolic state representations and (ii) causal processes for both endogenous actions and exogenous mechanisms. Each causal process models the time course of a stochastic cause-effect relation. We learn these world models from limited data via variational Bayesian inference combined with LLM proposals. Across five simulated tabletop robotics environments, the learned models enable fast planning that generalizes to held-out tasks with more objects and more complex goals, outperforming a range of baselines.
@article{arxiv.2509.26255,
title = {ExoPredicator: Learning Abstract Models of Dynamic Worlds for Robot Planning},
author = {Yichao Liang and Dat Nguyen and Cambridge Yang and Tianyang Li and Joshua B. Tenenbaum and Carl Edward Rasmussen and Adrian Weller and Zenna Tavares and Tom Silver and Kevin Ellis},
journal= {arXiv preprint arXiv:2509.26255},
year = {2026}
}
Comments
ICLR 2026. The last two authors contributed equally in co-advising